Electron band alignment between (100)InP and atomic-layer deposited Al 2O3

Article abstract of DOI:10.1063/1.3496039

Energy barriers at interfaces of (100)InP with atomic-layer deposited Al₂ O₃ are determined using internal photoemission of electrons. The barrier height between the top of the InP valence band and bottom of the alumina conduction band is found to be 4.05±0.10 eV corresponding to a conduction band offset of 2.7 eV. An interlayer associated with the oxidation of InP may result in a lower barrier for electron injection potentially leading to charge instability of the insulating stack. A wide-gap P-rich interlayer has a potential to reduce this degrading effect as compared to In-rich oxides.

Full text of DOI:10.1063/1.3496039

Electron band alignment between (100)InP and atomic-layer deposited Al 2 O 3
Hsing-Yi Chou, V. V. Afanas’ev, A. Stesmans, H. C. Lin, P. K. Hurley, and S. B. Newcomb
Citation: Applied Physics Letters 97, 132112 (2010); doi: 10.1063/1.3496039
View online: http://dx.doi.org/10.1063/1.3496039
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/97/13?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Electron band alignment at the interface of (100)GaSb with molecular-beam deposited Al 2 O 3
Appl. Phys. Lett. 98, 072102 (2011); 10.1063/1.3549878
Energy-band alignment of Al 2 O 3 and HfAlO gate dielectrics deposited by atomic layer deposition on 4 H –
SiC
Appl. Phys. Lett. 96, 042903 (2010); 10.1063/1.3291620
Energy barriers at interfaces between (100) In x Ga 1 x As ( 0 x 0.53 ) and atomic-layer deposited Al 2 O 3
and HfO 2
Appl. Phys. Lett. 94, 202110 (2009); 10.1063/1.3137187
Energy-band parameters of atomic layer deposited Al 2 O 3 and HfO 2 on In x Ga 1 x As
Appl. Phys. Lett. 94, 052106 (2009); 10.1063/1.3078399
Energy barriers at interfaces of (100)GaAs with atomic layer deposited Al 2 O 3 and HfO 2
Appl. Phys. Lett. 93, 212104 (2008); 10.1063/1.3021374
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
140.117.111.1 On: Fri, 15 Aug 2014 08:43:10

APPLIED PHYSICS LETTERS 97, 132112 ͑2010͒
Electron band alignment between „100…InP and atomic-layer
deposited Al2O3
Hsing-Yi Chou,1,a͒ V. V. Afanas’ev,1 A. Stesmans,1 H. C. Lin,2 P. K. Hurley,3 and
S. B. Newcomb4
1Department of Physics, University of Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium
2IMEC, 75 Kapeldreef, B-3001 Leuven, Belgium
3Tyndall National Institute, University College Cork, Lee Maltings, Prospect Row, Cork, Ireland
4Glebe Scientific Ltd., Newport, Tipperary, Ireland
͑Received 9 August 2010; accepted 12 September 2010; published online 30 September 2010͒
Energy barriers at interfaces of ͑100͒InP with atomic-layer deposited Al2O3 are determined using
internal photoemission of electrons. The barrier height between the top of the InP valence band and
bottom of the alumina conduction band is found to be 4.05Ϯ0.10 eV corresponding to a conduction
band offset of 2.7 eV. An interlayer associated with the oxidation of InP may result in a lower barrier
for electron injection potentially leading to charge instability of the insulating stack. A wide-gap
P-rich interlayer has a potential to reduce this degrading effect as compared to In-rich oxides.
© 2010 American Institute of Physics. ͓doi:10.1063/1.3496039͔
Like other AIIIBV semiconductors InP attracts consider-
able attention as possible high-mobility channel material for
next generations metal-insulator-semiconductor ͑MIS͒ elec-
tronic and optoelectronic devices. Recently, MIS transistors
employing InP and the atomic-layer-deposition ͑ALD͒
grown Al2O3 as gate insulator have been demonstrated.1–4
However, InP devices were reported to suffer from the
threshold voltage instability2 caused by charge injection and
trapping in the insulator. This issue naturally brings band
alignment at the InP/Al2O3 interface to attention of an ex-
perimentalist. Moreover, reliable knowledge of band align-
ment is needed for proper engineering of more complex
stacks that use InP as a barrier layer on top of the InxGa1−xAs
in quantum well channel structures.3,4 In the present work we
determine the electron band alignment between ͑100͒InP and
ALD Al2O3 using internal photoemission ͑IPE͒ of electrons.
In particular, it is found that oxidation of the InP surface
results in formation of a P-rich wide gap interlayer ͑IL͒,
quite in contrast to the earlier studied interfaces of arsenides
with Al2O3 exhibiting a narrow gap IL.5–8 This finding
makes the InP/Al2O3 interface particularly promising for ap-
plication in MIS devices.
the latter oxide to be P-rich as the electron scattering ampli-
9,10
tude increases with atomic number ͑Z͒ as Z3/2
.
Thus, a
weak contrast between the oxides of Al ͑Z=13͒ and P ͑Z
=15͒ is expected, while the absence of the contrast associ-
ated with In ͑Z=49͒ suggests that most of the In is probably
removed during ALD.
For IPE measurements, MIS capacitors were prepared by
evaporation of semitransparent ͑13-nm thick͒ Au or Al elec-
trodes of 0.5 mm2 area onto Al2O3. The IPE together with
photoconductivity ͑PC͒ measurements were performed at
room temperature in the photon energy ͑h␯͒ range 2.0–6.8
eV. The quantum yield ͑Y͒ was determined by normalizing
the measured photocurrent to the incident photon flux.11 An
example of the yield spectral curves is shown in Fig. 2 for
the n-type InP/Al2O3/Au sample with HCl predeposition
clean. Spectra measured under positive bias ͑open symbols͒
show field-independent features at EЈ=4.7 eV and E2
=5.1 eV, coinciding in energy with the 0excitation of direct
optical transitions between high-symmetry points in the Bril-
louin zone of InP, i.e., ⌫V8 →⌫7C and X5V→X3C, respectively.12
This observation leaves little doubt that the photocurrent ob-
Studied samples were prepared on single crystal ͑100͒
InP of n-type ͑S-doped͒ or p-type ͑Zn-doped͒ conductivity
with doping concentration of about 4ϫ1017 cm−3. Three
predeposition surface treatments were compared: wet etch in
HCl ͑0.1% water solution͒, the same etch followed by treat-
ment in ͑NH4͒2S from a 25% polysulfide water solution, and
UV/O3 oxidation. Next, a layer of Al2O3 ͑8 nm target thick-
ness͒ was deposited by ALD at 300 °C using Al͑CH3͒3 ͑the
first pulse͒ and H2O precursors. Transmission electron mi-
croscopy ͑TEM͒ cross-sectional images show almost no con-
trast between the IL and the Al2O3 layer. Nevertheless, the IL
can be visualized ͑Fig. 1͒ after prolonged sample exposure to
an electron beam causing alumina to crystallize, while the IL
remains amorphous. In this way the thickness of the IL,
which appears to be sensitive to the preclean, can be deter-
mined. The low contrast between Al2O3 and the IL indicates
FIG. 1. Cross-sectional TEM images of the InP/Al2O3 interface in samples
prepared using HCl treatment ͑a͒ and UV/O3 oxidation ͑b͒ prior to ALD.
The arrows mark the thicknesses of the Al2O3 layer and the IL grown be-
tween Al2O3 and InP, observed after crystallizing alumina by an electron
beam.
a͒
Electronic mail: hsingyi.chou@fys.kuleuven.be.
0003-6951/2010/97͑13͒/132112/3/$30.00
97, 132112-1
© 2010 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
140.117.111.1 On: Fri, 15 Aug 2014 08:43:10

132112-2
Chou et al.
Appl. Phys. Lett. 97, 132112 ͑2010͒
FIG. 2. IPE and PC yield as a function of photon energy measured on
InP/Al2O3 /Au samples, prepared using pre-ALD HCL etching, with the
applied bias varying from 1.0 to 3.0 V ͑open symbols͒ and from Ϫ1.0 to
Ϫ3.0 V ͑filled symbols͒. Vertical lines indicate energies of direct optical
transitions within the InP crystal.
served in the spectral range h␯=3.5–5.5 eV is due to IPE of
electrons from the InP valence band ͑VB͒ into the oxide
conduction band ͑CB͒. By contrast, no optical signatures of
InP are seen in the spectra taken under negative bias ͑filled
symbols͒ suggesting electron IPE from Au into Al2O3. In-
deed, replacement of Au with Al shifts the spectral threshold
by Ϸ1 eV to lower h␯ ͑spectra not shown͒. At h␯Ͼ6 eV the
photocurrent spectra for both orientations of the electric field
reveal the onset of intrinsic PC in alumina with spectral
threshold at Eg͑Al2O3͒=6.1Ϯ0.1 eV ͑cf. insert in Fig. 3͒ in
agreement with results for ALD Al2O3/Si.13 Noteworthy is
that in the UV/O3 oxidized ͑prior to ALD͒ sample with the
thickest IL a second PC threshold is found at 6.5Ϯ0.1 eV
apparently corresponding to band-to-band excitation within
the IL.
FIG. 3. ͑a͒ Determination of the IPE spectral thresholds in the pre-ALD
HCl-treated n-InP/Al2O3 /Au sample from the spectra shown in Fig. 2 using
Y
1/3-h␯ ͑Powell͒ plots. Vertical arrows indicate the inferred spectral thresh-
olds. The inset illustrates determination of the electron IPE threshold from
Au into Al2O3 ͑⌽Au͒ and the oxide band gap ͓Eg͑Al2O3͔͒ using the Y1/2-h␯
plot; ͑b͒ Comparison of the IPE spectra between samples prepared using
three different InP surface treatments. All curves are measured under +2 V
bias applied to Au electrode.
=3–4 eV, as evident from Fig. 2 ͑open symbols for positive
bias͒. This observation suggests that electrons need to tunnel
through some barrier to enter the CB of Al2O3. Consistent
with this hypothesis is that the IPE yield increases exponen-
tially with electric field as well, as both the barrier height and
the field enter together into the tunneling exponent ͓see, e.g.,
Eq. ͑1͒ in Ref. 15 or the Fowler–Nordheim expression͔.
To determine interface energy barriers, the spectral
thresholds were plotted as a function of the square root of the
average electric field in the oxide stack ͑the Schottky plot͒.
The field was calculated by subtracting the built-in voltage,
at which the IPE current appears, from the applied bias to
account for the metal-InP work function difference, and then
dividing the obtained value by the oxide thickness estimated
from TEM images. The results, compiled for all studied
samples in Fig. 4͑a͒, clearly show that the threshold ⌽1 is
barely sensitive to the initial InP surface preparation and,
therefore, can be associated with direct IPE into the CB of
Al2O3. By contrast, the lower threshold ⌽2 is highly sensi-
tive to the pre-ALD surface treatment and even to the con-
ductivity type of the InP substrate allowing us to associate it
with the IPE from InP into the CB of the IL. In Fig. 4͑a͒ we
also compare ⌽2 values obtained from the ⌼1/3-h␯ plot ͑᭿͒
to those obtained from the log plot shown in Fig. 2 by ex-
trapolating the yield to the sub-threshold background level
͑small filled squares͒.
To determine the IPE spectral thresholds from the data
shown in Fig. 2 for the HCl-treated n-InP/Al2O3/Au sample,
we plotted in Fig. 3͑a͒ the yield in ⌼1/3-h␯ ͑IPE from InP͒ or
⌼
1/2-h␯ ͑IPE from Au and Al͒ coordinates as suggested by
Powell.14 The IPE from Au ͓inset in Fig. 3͑a͔͒ obeys the
Fowler law yielding a threshold of ⌽Au=4.1 eV, in good
agreement with the value found in ͑100͒Si/Al2O3/Au
structures.13 By contrast, the IPE spectra from the VB of InP
cannot be described by a single threshold value. From the
part of the curve measured in the range 3.8Ͻh␯Ͻ4.5 eV
one can still find the threshold ⌽1, similar to ⌽Au, but the
origin of the low-energy “tail” stretching down to almost 3
eV and the corresponding threshold ⌽2 is not evident. To
shed some light on the source of the photocurrent in the
range 3.2Ͻh␯Ͻ4 eV, we compare in Fig. 3͑b͒ the ⌼1/3-h␯
plots of n-InP/Al2O3 samples prepared using different pre-
deposition surface treatments. Samples with HCl ͑ᮀ͒ and
HCl+͑NH4͒2S ͑᭝͒ preclean show very similar spectra, while
in the sample subjected to UV/O3 oxidation ͑᭺͒ the IPE is
significantly attenuated. This attenuation correlates with the
thicker IL ͑Fig. 1͒ suggesting scattering of excited electrons
inside the IL as the major yield reduction mechanism. To be
noted is that the IPE yield is reduced by the thicker IL over
the whole photon energy range covered, indicating InP as the
common source of photoelectrons. Further insight on the un-
usual behavior of the IPE from InP is provided by the obser-
Extrapolation to zero field gives consistent results
0
2
with average values ⌽01=4.05Ϯ0.10 eV and
⌽
vation of an exponential yield increase in the range h␯
=3.70Ϯ0.15 eV, thus even when using different threshold
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
140.117.111.1 On: Fri, 15 Aug 2014 08:43:10

132112-3
Chou et al.
Appl. Phys. Lett. 97, 132112 ͑2010͒
and even to 3.5 eV in anodic oxide on InP.20 These narrow-
gap phases may further reduce the IL-related barrier poten-
tially explaining the anomalously strong field-induced barrier
lowering in some samples as demonstrated ͑⌽2͒ in Fig. 4͑a͒.
Moreover, we found that at the ͑100͒InP/ALD HfO2 inter-
face the spectral threshold of electron IPE only slightly ex-
ceeds 2 eV ͑spectra not shown͒. This is likely caused by
In-rich IL formation and may explain the recently reported
charge instability of InP/HfO2 interfaces.21
In conclusion, we found that ALD Al2O3 provides a high
CB offset, of 2.7 eV, with ͑100͒InP. However, it is also ob-
served that the formation of an IL due to oxidation of semi-
conductor may reduce the barrier height for electrons signifi-
cantly, particularly if a high electric field is applied ͓cf. Fig.
4͑a͔͒. To avoid this undesirable effect, the concentration of
indium in the IL should be reduced; According to the results
P-rich oxide on InP provides a sufficiently large CB offset in
excess of 2 eV.
PH acknowledges the financial support of Science Foun-
dation Ireland under Grant No. SFI/09/IN.1/I2633.
1Y. Q. Wu, Y. Xuan, T. Shen, P. D. Ye, Z. Cheng, and A. Lochefeld, Appl.
Phys. Lett. 91, 022108 ͑2007͒.
2H. Zhao, D. Shahrjerdi, F. Zhu, M. Zhang, H.-S. Kim, I. Ok, J. H. Yum, S.
I. Park, S. K. Banerjee, and J. C. Lee, Appl. Phys. Lett. 92, 233508
͑2008͒.
3H. Zhao, Y.-T. Chen, J. H. Yum, Y. Wang, N. Goel, and J. C. Lee, Appl.
Phys. Lett. 94, 193502 ͑2009͒.
FIG. 4. ͑a͒ Schottky plot of the field dependent IPE thresholds measured for
differently prepared InP/Al2O3 interfaces. For the n-InP/Al2O3 sample with
pre-ALD HCl treatment the values of the lower barrier ⌽2 obtained from the
log͑Y͒-h␯ ͑Fig. 2͒ and Y1/3-h␯ ͑Fig. 3͒ plots are shown by small and large
filled squares, respectively. Lines illustrate the determination of average
zero-field barrier heights ⌽10 and ⌽02 between the top of the InP VB and the
CB bottom of Al2O3 and of the IL, respectively; ͑b͒ Energy band diagram of
the InP/Al2O3 /Au structure inferred from the measured barrier values. The
origin of the energy scale is at the top of the InP valence band.
4H. Zhao, Y.-T. Chen, J. H. Yum, Y. Wang, F. Zhou, F. Xue, and J. C. Lee,
Appl. Phys. Lett. 96, 102101 ͑2010͒.
5N. V. Nguyen, O. A. Kirillov, W. Jiang, W. Wang, J. S. Suehle, P. D. Ye,
Y. Xuan, N. Goel, K. W. Choi, W. Tsai, and S. Sayan, Appl. Phys. Lett.
93, 082105 ͑2008͒.
6V. V. Afanas’ev, M. Badylevich, A. Stesmans, G. Brammertz, A. Delabie,
S. Sionke, A. O’Mahony, I. M. Povey, M. E. Pemble, E. O’Connor, P. K.
Hurley, and S. B. Newcomb, Appl. Phys. Lett. 93, 212104 ͑2008͒.
7V. V. Afanas’ev, A. Stesmans, G. Brammertz, A. Delabie, S. Sionke, A.
O’Mahony, I. M. Povey, M. E. Pemble, E. O’Connor, P. K. Hurley, and S.
B. Newcomb, Appl. Phys. Lett. 94, 202110 ͑2009͒.
determination plots for ⌽2. With these values together with
the band gap width of the IL and Al2O3, and the measured
barrier between the Fermi level of Au and the Al2O3 CB,
we arrive at the electron band diagram shown in Fig. 4͑b͒.
The CB offsets can be estimated by subtracting the known
band gap of InP ͑1.35 eV at 300 K͒ from the measured
barriers: ⌬Ec͑InP/Al2O3͒=2.7Ϯ0.10 eV and ⌬Ec͑InP/IL͒
=2.35Ϯ0.15 eV. Here, we note that the former offset nearly
coincides with the theoretical prediction,16 while the second
one complies with the offset estimated for the InP/In͑PO3͒3
interface.17 Apparently then, the P-rich IL may be associated
with In͑PO3͒3 as the inferred IL band gap width of 6.5 eV is
close to the 6.9 eV gap of In͑PO3͒3 layers grown on InP.17,18
The above results suggest that, similarly to the previ-
ously studied interfaces of arsenides,6–8 the IL plays an im-
portant role in electron injection processes as it provides the
lowest barrier for electrons in InP. Potentially, oxidation of
InP may result in oxides with an even narrower gap than that
of In͑PO3͒3: In InPO4 the gap narrows to 4.5 eV ͑Ref. 19͒
8N. V. Nguyen, M. Xu, O. A. Kirillov, P. D. Ye, C. Wang, K. Cheung, and
J. S. Suehle, Appl. Phys. Lett. 96, 052107 ͑2010͒.
9S. J. Pennycook, Ultramicroscopy 30, 58 ͑1989͒.
10D. B. Williams and C. B. Carter, Transmission Electron Microscopy ͑Ple-
num, New York, 1996͒, p. 353.
11V. V. Afanas’ev and A. Stesmans, J. Appl. Phys. 102, 081301 ͑2007͒.
12V. I. Gavrilenko, A. M. Grekhov, D. V. Korbutyak, and V. G. Litovchenko,
Optical Properties of Semiconductors ͑Naukova Dumka, Kiev, 1987͒.
13V. V. Afanas’ev, M. Houssa, A. Stesmans, and M. M. Heyns, J. Appl.
Phys. 91, 3079 ͑2002͒.
14R. J. Powell, J. Appl. Phys. 41, 2424 ͑1970͒.
15G. B. Alers, K. S. Krisch, D. Monroe, B. E. Weir, and A. M. Chang, Appl.
Phys. Lett. 69, 2885 ͑1996͒.
16J. Robertson and B. Falabretti, J. Appl. Phys. 100, 014111 ͑2006͒.
17Y. Robach, M. P. Besland, J. Joseph, G. Hollinger, P. Viktorovitch, P.
Ferret, M. Pitavel, A. Falcou, and G. Post, J. Appl. Phys. 71, 2981 ͑1992͒.
18G. Lévêque and Y. Villachon-Renard, Appl. Opt. 29, 3207 ͑1990͒.
19J. F. Wager, C. W. Wilmsen, and L. L. Kazmerski, Appl. Phys. Lett. 42,
589 ͑1983͒.
20A. A. Studna and G. L. Gualtieri, Appl. Phys. Lett. 39, 965 ͑1981͒.
21Y.-T. Chen, H. Zhao, Y. Wang, F. Xue, F. Zhou, and J. C. Lee, Appl. Phys.
Lett. 96, 253502 ͑2010͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
140.117.111.1 On: Fri, 15 Aug 2014 08:43:10